Statistical Methods in Finance 2025

Estimating Large Dense Covariance Matrices under Changing Market Conditions

By:Kathy Ensor
Center for Computational Finance and Economic Systems, Rice University

The key to risk metrics in corporate finance is the covariance between assets and market sectors. It may be necessary to estimate up to 10,000 covariances quickly. The Gaussian copula approach to covariance estimation is computationally efficient, but is it robust? We will explore estimating covariances using time series methods and adapt these approaches to rapidly changing market conditions. Our methodology will offer short-term covariance forecasting and long-term covariance scenario analysis. Computational speed will be a central focus.